Derivations of Upper Triangular Matrix Rings

S6nia P. Coelho and C. Polcino Milies

Instituto de Matem&ca e Estatktica
Universidade de SZo Paul0
Caixa Postal 20.570-Ag. lguatemi
Go Paula, Bra.&

Submitted by Graciano de Oliveira

ABSTRACT

We give a description of the derivations in T,(R), the ring of upper triangular

matrices over a ring R, assuming only the existence of an identity element. We show
that every derivation is the sum of an inner derivation and another one induced from
R.

1. INTRODUCTION

Let R be a ring with unity, and denote by T,,(R) the ring of upper
triangular n X n matrices over R. The group of automorphisms of this ring,
given certain restrictions on R, has been the subject of several recent papers
(see [I, 2, 51). In this note we shall give a description of the derivations in
T,(R) assuming no restriction on R other than the existence of an identity
element. We shall show that they are obtained in a very natural way.
First, note that if 6 : R + R is a derivation of R, then the map 8 : T,(R)
--f T,(R) defined by @x,~) = (6(xij)) V(xij> E T,(R) is a derivation. Also, if
A E T,(R), we shall denote by d, the inner derivation induced by A. In
what follows, we shall show that all derivations of T,(R) arise from these;
namely, we prove the following.

THEOREM. Let R be a ring with unity and let d : T,(R) -+ T,(R) be a

derivation. Then there exists a derivation 6 : R -+ R and a matrix A E T,,(R)
such that d = S + d,.

LINEAR ALGEBRA AND ITS APPLICATlONS 187: 263-267 (1993) 263

0 Elsevier Science Publishing Co., Inc., 1993

655 Avenue of the Americas, New York, NY 10010 0024.3795/93/$6.00
264 SbNIA P. COELHO AND C. POLCINO MILIES

A similar result for full matrix rings appears in [6], and the special case
where R is an algebra over a field, with char(R) # 2,3 and n > 2, is given in
[4]. The case of upper triangular matrix rings over a simple algebra finite
dimensional over its center appears in [5].
Our methods are different in that our proof is very simple and construc-
tive in nature: in Lemma 1 we show how the matrix A can be fully computed
from the values of d on the usual matrix units eij E Z’,(R). In Lemma 2 we
also show how to obtain 6 from d. In fact, these techniques also apply to the
ring M,(R), so they give an elementary proof for the description of the
derivations in this ring.

2. PROOF OF THE THEOREM

Notice that a matrix unit e,J belongs to T,,(R) provided that i < j. Since
ekk is an idempotent, we have that d(ek,) = ekkd(ekk) + d(ekk)ekk. Thus, if
we write d(e,,> = CiGlxjl(k)eij, we have that

d(ekk) = Cxij(k)eij = C XkJ(k)ekj + C rik(k)e,k>

i<j jrk i<k

which shows readily that r,,(k) = 0 and that xii(k) = 0 if i <j are both
different from k.
Given a matrix X E T,(R), we shall d enote by [Xl,, the i, j entry of this
matrix. If i <j, we define njj = -[e,,d(eii)ejj]jj, i.e. nzj = -xii(i). Also we
set aij = 0 if i > j.

LEMMA 1. Let A be n matrix with entries chosen us above and arbitrary

diagonal entries. Set h = d - d,; then:

(i) h(ekk) = 0, 1 < k =Gn.

(ii) Furthermore, if we set ujj = -[d(e,i>l,j, 1 <j < n, then h(eij) = 0
fori <j.

Proof. (i): We compute

By our definition - uk = xk .( k ). Also, notice that if e, f E T,,(R) are orthog-

onal idempotents, we have that en(f) + d(e)f = 0, which implies e&f>f =
DERIVATIONS OF MATRIX RINGS 265

-ed(e)f. Hence, in particular, ei,&ekk)ekk = -e,,d(e,,)ekk, SO if i < k

then uik = --r,,(i). Hence d,(ekk) = d(ekk), 1 < k < n, and 6) follows.
(ii): We shall show first that h(e,j) = 0, 1 <j < n. Since elj = e,,e,jejj,
using (i) we obtain

= ell(d( e,j) - AelI + e,jA)ejl

= -',jjelj -
a,,e,j + ujje,j = 0.

Now we prove that h(eij) = 0, 1 < i <j. In fact, since eli = el,ejj, we
see that

0 = h(e,j) = e,,h(eij) = e,,d(eij) - e,,Aeij + e,,ezjA.

Hence

[ d( ei,j)] ‘j = n,, - ujj.

Finally, we have that eij = eiieijejJ, so we can write

h( ejj) = e,,h( eij)ejj

= e,,d( e,j)e,ij - e,, Aeii + eij Aejj

= [ d( e,,)] ,jeij - aiieij + aljjeij = 0.

From now on, we shall assume that the elements on the diagonal of A are
chosen as in (ii) above, and hence that h(e,j)
= 0, i < j.

LEMMA2. There exists a derivation 6 : R -+ R such that h(rZ) = S(r)Z

Vr E R.

Proof. Set r E R. Notice that

h( reii) = h( rZ)e,, = e,,h( rZ), l<i$n,

266 SGNIA P. COELHO AND C. POLCINO MILIES

and for every pair i < j, we have that

h( reij) = h( reiieij) = h( reii)ejj = h( eijre,ij) = ejjh( rejl).

. a diagonal matrix and hence, by

The first set of equations shows that h(rZ) 1s
the second set of equations, h(rZ) = r’l.
It is now easy to see that the map 6 : R + R given by r * r’ is a
derivation of R. n

proof of the Theorem. Set (Y = (rij)i ~ j E T,,(R). We have that

(d - dA)( a) = h( cx) = h( xrijZeij)

i<j

= CG(rij)Zeij = C6(rij)eij = S(a).

i<j isj

Hence, d(a) = d,(a) + 6(a) VCYE T,(R).

It may be interesting to note that the derivation 6 and the matrix A are
unique up to inner derivations.

PROPOSITION. Let d : T,,(R) -+ T,,(R) be a derivation, and let A, 6 be as

in the theorem above. Zf A’ E T,(R) and 6 ’ : R + R are such that d = d,,
+ s’, then there exists an element a E R such that 6 ’ = 6 - 6, and A’ = A
+ aZ (so d,, = d, + d,,).

Proof. We have that d,, - d, = 8 - s’, so, computing for an arbitrary

elementary matrix ejj, we obtain

(A’ - A)eij - eij( A’ - A) = 0,

i.e.,

Hence, aik = ahk if h # k, and ati - aji = a;3 - aj..

If we set j = 1, we obtain ati = alI + aij, so 1.I we denote a = a’,,, we
have that A’ = A + al.
DERIVATIONS OF MATRIX RINGS 267

If we choose r E R, we can compute

(CIA’ - dA)(?-I) = (8 - 6’)(rI),

which gives

(a?” - t-u)1 = [S(r) - S’(r)]I,

so S’ = 6 - s,. n

We note that the derivation rl : T,(R) + T,(R) is inner if and only if the
corresponding derivation 6 : R --+ R is inner. This fact, together with the
well-known theorem of Skolem and Noether, readily gives:

THEOREM. Let R he afinite dimensional centrul simple algebra Then all

linear derivations (IfT,(R) are inner.

As we mentioned before, the same techniques also prove the following.

THEOREM [4, 61. Let R be a ring with unity, and let d : M,(R) + M,(R)
be a derivation. Then there exists a derivation 6 : R -+ R and a matrix
A E M,,(R) such that d = 8 + d,.

REFERENCES

1 G. P. Barker, Automorphism groups of algebras of triangular matrices, Linear

Algebra A&. 121:207-215 (1989).
2 G. P. Barker and T. P. Kezlan, Automorphisms of algebras of upper triangular
matrices, Arch. Math. 55:38-43 (1990).
3 J. Bergen, I. N. Herstein, and C. Lanski, Derivations with invertible values,
Canad. J. Math. 35(2):300-310 (1983).
4 G. M. Benkart and J. M. @born, Derivations and automorphisms of nonassocia-
tive matrix algebras, Trans. Anzer. Alath. Sot. 263:411-430 (1981).
5 S. Jondrup, Automorphisms of upper triangular matrix rings, Arch. Math.
49:497-502 (1987).
6 D. Mathis, Differential polynomial rings and Morita equivalence, Comm. Algebra
lO:ZOOl-2017 (1982).

Received 16 july 1991; final munuscript accepted 15 July 1992